Member_2_231077
asked on
Daisy-chained SAS power-up order
I have a DL380 G5 with an additional P800 SAS controller in a PCIe slot, it has two external 4-lane SAS ports. On one of the external ports of the P800 I have an MSA70 disk shelf [with an input and output SAS port] and connected to that MSA70 there is another MSA70. Let us call them Server, diskbox1 and diskbox2.
Question: In what order should I power them on?
Before someone suggests I eliminate the 'peripheral of a peripheral' problem because I have two external channels I have four MSA60s daisy-chained together on the other external SAS port and the same question applies to that chain.
Second question: What is it that ultimately controls this recommended power up order?
Question: In what order should I power them on?
Before someone suggests I eliminate the 'peripheral of a peripheral' problem because I have two external channels I have four MSA60s daisy-chained together on the other external SAS port and the same question applies to that chain.
Second question: What is it that ultimately controls this recommended power up order?
ASKER
But isn't it a bit like parallel SCSI where the peripherals must be powered on first?
The exp-boxes power up first and then server. Let them spin up properly. If I remember right the SAS enclsoures do not need any "BOX ID" to be assigned to them as in some other protocols.
Powering up the exp-drives first will prevent loads of potential trouble for ya ;)
Powering up the exp-drives first will prevent loads of potential trouble for ya ;)
ASKER
Yes, server last.
But which external disk box to power up first when there are two or more external enclosures daisy-chained together? It's not as simple as a shared bus since there is a fan-out chip in the first enclosure that sends half the commands to the second enclosure and passes on the other half of the commands to it's own disks?
(not sure what exp means)
But which external disk box to power up first when there are two or more external enclosures daisy-chained together? It's not as simple as a shared bus since there is a fan-out chip in the first enclosure that sends half the commands to the second enclosure and passes on the other half of the commands to it's own disks?
(not sure what exp means)
Exp=Expansion box meaning the MSA
You can power the MSA boxes in any order. It does not affect the bus itself. I would say the OS is more sensitive than the hardware in this case.
If you would power up the MSA boxes after the servers you would most likely (not always) see the disks after a rescan. I have played with this and sometimes it can give you some interesting "effects"
Another thing which is obvious... if you have RAID set spanning over several shelves, a wrong startup procedure would break the set
You you really want to make sure there are no issues whatsoever I would power up in this order:
diskbox1 (1st in chain)
diskbox2 (because connected to 1)
etc...
Server(s)
You can power the MSA boxes in any order. It does not affect the bus itself. I would say the OS is more sensitive than the hardware in this case.
If you would power up the MSA boxes after the servers you would most likely (not always) see the disks after a rescan. I have played with this and sometimes it can give you some interesting "effects"
Another thing which is obvious... if you have RAID set spanning over several shelves, a wrong startup procedure would break the set
You you really want to make sure there are no issues whatsoever I would power up in this order:
diskbox1 (1st in chain)
diskbox2 (because connected to 1)
etc...
Server(s)
ASKER
It seems the behaviour changes depending on which rev of SAS Management Protocol is in use, searching HP's web site is hopeless, there isn't any mention of Broadcast (Change) which is what I think the MSAs are doing wrong. This is rather annoying since HP obviously know what is going on since they have a related patent. (www.freepatentsonline.com/20060101171.html)
The boxes with firmware 1.50 work exactly as you suggest - power on the furthest one first, 1.42 firmware seem more likely to power on clean if done the other way around and with mixed versions I was getting wierd behaviour. Unfortunately I've logged the problem with HP support who say there is no firmware in the MSAs!
Rescan (discovery) ought to find everything and the controller does this when server is booting up but it appears the expander has disabled the out port in some cases.
The boxes with firmware 1.50 work exactly as you suggest - power on the furthest one first, 1.42 firmware seem more likely to power on clean if done the other way around and with mixed versions I was getting wierd behaviour. Unfortunately I've logged the problem with HP support who say there is no firmware in the MSAs!
Rescan (discovery) ought to find everything and the controller does this when server is booting up but it appears the expander has disabled the out port in some cases.
ASKER
I think I'm getting near the truth, http://t10.org/ftp/t10/document.06/06-459r0.pdf
ASKER CERTIFIED SOLUTION
membership
Create a free account to see this answer
Signing up is free and takes 30 seconds. No credit card required.
ASKER
Indeed, it ought to work hot adding and I was wrong at first since the order shouldn't matter. It's got to be a bug in HP's code.
Given that these are external cases ,you can set them up both the same as there will not be a massive current in rush on one of them.